2000 - Intrafractional Motion Analysis of Pancreatic Tumors and Organs-at-Risk during MRgART Using Optical Flow Calculation

Purpose/Objective(s): Magnetic resonance-guided adaptive radiotherapy (MRgART) facilitates real-time acquisition of 2D cine MR images during beam delivery. However, no prior studies have evaluated the intrafractional anatomical variations of pancreatic tumors and surrounding organs-at-risk (OARs), including the stomach, duodenum, small bowel, and large bowel, during MRgART. This study aimed to quantify the motion range of these structures and analyze their correlation using optical flow calculation.

Materials/Methods: This study included 18 consecutive patients with pancreatic cancer who underwent MRgART between September 2022 and September 2024. All patients were treated with a Unity MR-LINAC with abdominal compression. During each fraction, single-slice 2D cine MR images were acquired every 600 ms in the axial, coronal, and sagittal planes at the isocenter during beam delivery. The first coronal frame for each fraction was used to delineate the tumor and OARs, including the stomach, duodenum, small bowel, and large bowel. Displacements in the superior-inferior (SI) direction were measured every 30 seconds within the overlapping regions of each delineated structure and a 10×10 cm square region centered on the isocenter using the Farnebäck optical flow method. The motion range in each direction was defined as the 2.5th–97.5th percentile of the displacements to exclude outliers. The root mean square error (RMSE) between the positions of the tumor and the OARs was calculated. Statistical differences in motion range and RMSE were assessed using the Wilcoxon signed rank test with Holm-Bonferroni correction (p < 0.05).

Results: The median acquisition time for 2D cine MR images was 616 s (range, 386–973 s). The median SI motion range of the tumor was 5.7 mm (IQR, 4.5–7.5 mm), which was smaller than that of all OARs (p < 0.05). Among the OARs, the large bowel (median, 10.1 mm; IQR, 7.8–13.1 mm) and small bowel (median, 9.3 mm; IQR, 7.6–11.7 mm) exhibited larger motion ranges compared to the duodenum (median, 7.4 mm; IQR, 5.8–9.6 mm) and stomach (median, 7.3 mm; IQR, 6.1–8.8 mm) (p < 0.05). The RMSE of the displacement between the tumor and the OARs was the highest for the large bowel (median, 1.8 mm; IQR, 1.3–2.5 mm), followed by the small bowel (median, 1.4 mm; IQR, 1.1–2.0 mm), while the duodenum (median, 1.0 mm; IQR, 0.7–1.3 mm) and the stomach (median, 0.9 mm; IQR, 0.6–1.2 mm) showed significantly smaller RMSE values (p < 0.05).

Conclusion: Our optical flow-guided intrafractional motion analysis revealed that SI motion range of pancreatic tumors was smaller than that of the surrounding OARs, including the stomach and duodenum, despite their proximity to the tumors. These findings highlight the importance of determining appropriate margins that account for intrafractional anatomical variations in both the tumors and the surrounding OARs during real-time monitoring in pancreatic MRgART.